Population Genetics of Sporophytic Self- Incompatibility in Senecio Squalidus L

Home , Senecio cambrensis, Senecio squalidus

Population genetics of sporophytic self- incompatibility in Senecio squalidus L. (Asteraceae) II: a spatial autocorrelation approach to determining mating behaviour in the presence of low S allele diversity

AC Brennan1, SA Harris1 and SJ Hiscock2 1Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK; 2School of Biological Sciences, University of Bristol, Woodland Road, Bristol BS8 1UG, UK

We recently estimated that as few as six S alleles represent mating events regularly occur at all the distance classes the extent of S locus diversity in a British population of the examined from 60 to 480 m throughout the entire sample self-incompatible (SI) coloniser Senecio squalidus (Oxford population. Less SA is observed for S locus data than for Ragwort). Despite the predicted constraints to mating allozyme data in accordance with the hypothesis that SSI imposed by such a low number of S alleles, S. squalidus and low diversity at the S locus are driving these large-scale maintains a strong sporophytic self-incompatibility (SSI) mating events. The limited population structure at small system and there is no evidence for a breakdown of SSI or distances of 60 m and less observed for SA analysis of the any obvious negative reproductive consequences for this Me-2 locus and by F-statistics for all the allozyme data, is highly successful coloniser. The present paper assesses evidence of some local relatedness due to limited seed and mating behaviour in an Oxford S. squalidus population pollen dispersal in S. squalidus. However, the overall through observations of its effect on spatial patterns of impression of mating dynamics in this S. squalidus popula- genetic diversity and thus the extent to which it is responsible tion is that of ample potential mating opportunities with many for ameliorating the potentially detrimental reproductive individuals at large population scales, indicating that repro- consequences of low S allele diversity in British S. squalidus. ductive success is not seriously affected by few S alleles A spatial autocorrelation (SA) treatment of S locus and available for mating interactions. allozyme polymorphism data for four loci indicates that Heredity (2003) 91, 502–509. doi:10.1038/sj.hdy.6800315

Keywords: sporophytic self-incompatibility; Senecio; S allele; mating behaviour; spatial autocorrelation analysis

Introduction the species during its introduction and subsequent colonisation of Britain over the past 300 years (Harris, Senecio squalidus L. (Oxford Ragwort, Asteraceae) main- 2002). tains a strong sporophytic self-incompatibility (SSI) SI species at evolutionary equilibrium characteristi- system governed by a single locus (S) (Hiscock, cally maintain high numbers of S alleles, through the 2000a,b). We previously investigated the number of S interaction of mutation, drift, and negative frequency- alleles in an Oxford population of S. squalidus and dependent selection, (Richman et al, 1996). Therefore, it identified just six S alleles (Brennan et al, 2002). Such a follows that SSI species with few S alleles may not be low estimate of S allele number is unusual for a large mating with optimal efficiency and may experience natural SSI population because estimates from other adverse fitness consequences. The threat to reproductive species have shown that typically 20–40 S alleles are assurance caused by low S allele number can be maintained in natural populations with SSI (Lawrence, quantified in terms of mate availability (MA), the 2000). High numbers of S alleles are usually a feature of proportion of compatible mates in SI populations (Byers natural self-incompatible (SI) populations since S alleles and Meagher, 1992). Reduced seed set and complete are subject to strong negative frequency-dependent reproductive failure due to low S allele number has been selection (Wright, 1939). It is likely that low S allele demonstrated in the threatened and rare Asteraceous diversity in British populations of S. squalidus is a species, Hymenoxys acualis var glabra (DeMauro, 1993), consequence of a population bottleneck experienced by and Aster furcatus (Reinartz and Les, 1994). Similarly, low S allele number has been cited as a reason for the relationship between seed set and population size in Correspondence: SJ Hiscock, School of Biological Sciences, University of Bristol, Woodland Road, Bristol BS8 1UG, UK. other Asteraceae, Eupatorium resinosum (Byers and E-mail: [email protected] Meagher, 1992) and Rutidosis leptorrhynchoides (Young Received 19 August 2002; accepted 4 April 2003 et al, 2000). The findings of these studies contrast Population genetics of SSI in Scenecio squalidus AC Brennan et al 503 markedly with the apparent success of S. squalidus in substructure develops whenever there are limitations to establishing itself and rapidly colonising available gene flow through pollen or seed dispersal, unless habitats in Britain following its introduction (reviewed disrupted by other processes, such as unifying selection in Harris, 2002). In other SSI species that have experi- or extinction–colonisation dynamics (Sokal and Warten- enced population bottleneck events, such as A. furcatus berg, 1983; Epperson, 1990). SA analysis can contribute to (Asteraceae), the negative reproductive consequences of determining the important processes shaping the dis- low S allele numbers have been offset by a breakdown of tribution of genetic diversity within populations, includ- SSI leading to self-compatibility (SC), where self-seed set ing levels of gene flow, which can be inferred through is possible (Reinartz and Les, 1994). SC may readily simulation studies (Epperson, 1995; Epperson and Li, evolve from SI under altered selective conditions since 1997; Epperson et al, 1999). Mating systems, such as SI, any mutation that disrupts the incompatibility of pollen tend to increase gene flow through pollen by increasing recognition or stigma response factors leads to self-fertile the scale of mating events when S allele-mediated individuals (Levin, 1996). However, the hypothesis that SSI incompatibilities are more probable between close may be in the process of breaking down to SC in S. relatives (Levin and Kerster, 1974; Levin, 1989). squalidus can be rejected since an effective SSI system The study of local population substructure of genetic operates in most wild-sampled and experimentally diversity in British S. squalidus is of particular interest in crossed individuals (Hiscock, 2000a,b; Brennan et al, the context of understanding its SSI mating system. This 2002). Thus, one or more alternative features of the life is because population substructure represents the net history of S. squalidus must be maintaining effective effect of the interacting processes of enhanced popula- reproduction within the framework of fully functional SSI. tion gene flow due to S-mediated incompatibilities Possible factors maintaining effective reproduction in coupled with potential mate availability limitations S. squalidus were discussed at length by Brennan et al imposed by very few S alleles. Additionally, this study (2002) and can be summarised as follows: (i) inherent represents a further development of the applications of features of the reproductive biology of S. squalidus; (ii) SA analysis to biological systems, since it is the first time increased dominance interactions between existing S that the mating dynamics of an SI species has been alleles; (iii) evolution at SI modifier loci other than the S investigated directly using S locus data. We would locus; and (iv) introgression of S alleles from related expect to observe less spatial autocorrelation for S locus Senecio species. Of the genetic factors considered, data integrally associated with mating dynamics and preliminary evidence suggests that increased dominance under strong negative frequency-dependent selection interactions between S alleles and the activity of modifier both temporally and spatially relative to selectively loci probably function as simple solutions to increasing neutral allozyme loci. Thus, a combined SA analysis MA, while maintaining a strong SI system in S. squalidus. including both S locus and allozyme data provides a These possibilities are currently under investigation in unique opportunity to test this hypothesis and poten- further British and Sicilian S. squalidus populations. This tially examine the effect that SSI mating dynamics has on paper will focus specifically on the role of the reproduc- other genetic diversity. tive biology of S. squalidus in maintaining effective reproduction within the context of a functional SSI Materials and methods system. A spatial autocorrelation (SA) methodology has been Plants applied to study the reproductive biology of S. squalidus S. squalidus by determining the spatial structure of genetic markers A total of 24 seedlings (Ox1–Ox27, excluding Ox6, Ox7, and Ox12 that died) were collected from a in an Oxford population. SA analysis was used rather than a more traditional population genetic-based population growing within a 0.5 km radius of Oxford railway station (OS grid ref.: SP505064) in March 1999 F-statistic methodology because SA methodologies are et al ideal for investigating small-scale spatial patterns of (Brennan , 2002; Figure 1). The positions of sampled plants were recorded on a 1:25 000 scale OS map, and genetic diversity at the level of individual populations. With SA analysis, data for individuals’ positions and later converted to coordinate data to the nearest 5 m relative to an arbitrary reference point. Seedlings were genotypes are used, whereas F analyses deal only with gene and genotype frequencies of groups of individuals grown and maintained in a glasshouse as described in Brennan et al (2002). necessitating arbitrary or assumptive a priori subdivi- sions (Bacilieri et al, 1994; Epperson and Li, 1996). Thus, SA is a highly sensitive measure of population spatial S genotyping structure even for relatively small sample sizes (Epper- A dominant S allele was identified for each plant using a son, 1993, 1995; Epperson and Li, 1996). full diallele cross and analysis of fruit-set phenotypes as SA has a long history of application to questions of previously described (Brennan et al, 2002). One indivi- biological interest (Sokal and Oden, 1978), while recent dual (Ox10) was found to be SC, based on intermediate syntheses of population genetic theory with popularly levels of self-seed set. Thus, an S allele could not be used SA statistics (eg Moran’s I) make SA an attractive assigned to this plant, which was omitted from further alternative to traditional F analyses (Hardy and Veke- analyses of S genotypes. mans, 1997). Here we have used SA analysis to gain a better understanding of within-population SSI-mediated Multilocus allozyme genotyping mating dynamics in S. squalidus. Young leaf tissue samples were prepared using the Identification of patterns of population substructure is methods of Ashton (1990). Proteins were separated by a powerful approach for determining species-mating electrophoresis, on 14% starch gels in a continuous Tris- dynamics. Theoretical studies show that population citrate buffer (pH 8.3) (Ashton, 1990; Glover and Abbott,

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 504 variation was confirmed in test crosses, where the allozyme genotypes of six Oxford (Ox) and five Glasgow (Gw) parental plants and up to 34 of their cross- progeny were determined and these genotype frequencies compared with expected frequencies using standard w2 tests (Tables 1 and 2). Allozyme genotype data were subjected to standard population genetic analysis of linkage, Hardy–Weinberg equilibrium and intra population inbreeding coefficient, Fis, using Genepop v.3.1d (Raymond and Rousset, 1995) and Fstat v.2.9.3 (Goudet, 1995).

Spatial autocorrelation analysis Spatial autocorrelation analysis of polymorphic allozyme genotype and transformed S locus data was performed using the program SGS v.1.0 (Degen et al, 2001). Moran’s I, a measure of spatial structure of genetic diversity at a single diallelic locus for sampled individuals within arbitrary distance classes, was calculated as follows (Sokal and Oden, 1978; Sokal and Wartenberg, 1983): P P n n n w ða À aÞða À aÞ i¼1 Pj6¼i ij i j I ¼ n 2 ð1Þ W i¼1 ðai À aÞ

where n is the total number of samples; wij ¼ 1or0, dependent on whether individuals i and j both belong to the same spatial interval or not; ai ¼ 1, 0.5, or 0, dependent on whether the ith individual contains 2, 1 or no copies of allele a, and a the mean value of ai over all Figure 1 Outline map of Oxford Railway station area showing n individuals. individual sample locations of S. squalidus Ox population. The null expectation for Moran’s I value in the absence of SA is equal to –1/(nÀ1) (Sokal and Wartenberg, 1983), which for this study is –0.04 (n ¼ 24). A Monte Carlo 1995). Gels were cut into 1 mm thick slices and stained permutation procedure of 1000 random redistributions of for the following enzymes (Wendel and Weeden, 1990): genotypes over coordinates was carried out on each aspartate aminotransferase (AAT, EC 2.6.1.1); aconitase measured Moran’s I statistic to determine their 95% b b À (ACO, EC 4.2.1.3); -esterase ( -EST, EC 3.1.1. ); confidence intervals and probability values according to glutamate dehydrogenase (GDH, EC 1.4.1.2); isocitrate Manly (1997), also using SGS v.1.0 (Degen et al, 2001). dehydrogenase (IDH, EC 1.1.1.42); malate dehydrogen- For each data set of genotypes considered, the null ase (MDH, EC 1.1.1.37); malic enzyme (ME, EC 1.1.1.40); expectation for Moran’s I, the measured Moran’s I 6-phosphogluconate dehydrogenase, (6-PGD, EC values, and their associated 95% confidence intervals 1.1.1.44), phosphoglucoisomerase (PGI, EC 5.3.1.9); and were plotted against distance to form correlograms or phosphoglucomutase (PGM, EC 5.4.2.2). The different graphical representations of SA. The overall 95% loci for each enzyme system and the different allozymes significance of each correlogram was evaluated by at each locus were identified. Loci were labelled calculating the probability of obtaining a correlogram numerically in order of decreasing migration distance more extreme than the one observed according to Sˇida´k’s from the cathode, while alleles were similarly labelled test, (Oden, 1984) as shown alphabetically to provide a multilocus allozyme geno- k type for each individual. The genetics of electrophoretic pðcÞ ¼ 1 Àð1 À mÞ ð2Þ

Table 1 Allozyme genotypes of parent plants used in test crosses of allozyme loci inheritance

Allozyme Parent plant crosspair

Ox5 Gw3 Ox17 Gw7 Ox20 Gw12 Ox25 Gw2 Ox26 Gw18 Ox10a Ox10a

Aat-1 bb ab bb ab bb aa bb bb bb ab aa aa Aco-1 bb ab ab ab ab ab bb ab ab bb bb bb best-1 ab bb bb bb bb ab bb bb bb aa aa aa Me-2 aa aa ab ab aa bb ab ab ab ab ab ab

Ox refers to parental plants chosen from the Oxford population described in the Materials and methods, while Gw refers to parental plants chosen from a separate Glasgow population. aProgeny array derived from spontaneous self-seed set in the SC individual Ox10.

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 505 Table 2 Progeny allozyme genotype frequencies and w2 tests for standard Mendelian inheritance.

Allozyme Genotype Parent plant crosspair

Ox5 Gw3 Ox17 Gw7 Ox20 Gw12 Ox25 Gw2 Ox26 Gw18 Ox10 Ox10

Progeny genotype frequencies

Aat-1 aa 000 0 029 ab 13 NS 14 NS 21 0 5 NS 0 bb 10 NS 8 NS 0 30 9 NS 0 Aco-1 aa 03NS4NS0 0 0 ab 8 NS 13 NS 9 NS 17 NS 7 NS 0 bb 9 NS 5 NS 7 NS 13 NS 6 NS 29 best-1 aa ÂÂÂ Â 010 ab ÂÂÂ Â 80 bb ÂÂÂ Â 00 Me-2 aa 11 9NS 0 9NS 2*NS 6NS ab 0 11NS 27 17NS 7*NS 10NS bb 0 6 NS 0 4 NS 5* NS 5 NS

NS=nonsignificant w2 test result for observed genotype frequencies. *NS=nonsignificant w2 test result for observed genotype frequencies when rarest genotype frequencies have been summed to expected frequency X5 and degrees of freedom modified accordingly. Â =allozyme locus not scored for this progeny array.

where p(c) is the probability of observed correlogram; m Table 3 Allozyme and S locus genotypes of Oxford sample the minimum individual probability value observed; and S. squalidus individuals the k the number of distance classes. The sample individual coordinate data were divided Plant Identified Aat-1 Aco-1 b-Est-1 Me-2 number S allele genotypes genotypes genotypes genotypes into eight consecutive distance classes of 60 m (0–480 m), which was the minimum distance class size that could be Ox1 S1 bb bb ab bb tested in order that the number of pairs of data points in Ox2 S2 ab bb bb bb each distance class averaged 27.8 (SE 2.6)-close to the Ox3 S2 bb aa bb bb minimum statistical sample size requirement for Moran’s Ox4 S3 ab bb bb ab I (Waser and Mitchell, 1990). Ox5 S1 bb bb ab aa Polymorphic allozyme loci with minimum allele Ox8 S4 bb ab ab ab frequencies of at least 0.05 were included in the SA Ox9 S1 ab bb bb aa Ox10 ?aabbbbab analysis. When loci were diallelic, both alleles provide Ox11 S1 bb bb bb ab identical SA information, so each polymorphic allozyme Ox13 S4 ab ab ab ab locus yielded one correlogram upon SA analysis. Ox14 S4 bb bb bb ab Combined Moran’s I values for all the allozyme loci Ox15 S4 ab ab bb ab combined were calculated by summing the numerator Ox16 S1 aa ab bb ab and denominator of equation (1) over the number of Ox17 S5 bb ab bb ab Ox18 S4 bb bb bb bb alleles tested (Streiff et al, 1998). Ox19 S6 bb bb bb ab The same analysis was performed on the dominant S Ox20 S3 bb ab bb aa genotype data coded as two alternative ‘genotype’ Ox21 S4 bb bb bb bb formats, Ssame and Sdiff. The Ssame data set was derived Ox22 S4 ab aa bb ab by coding the unidentified S allele of each individual as Ox23 S4 ab bb bb ab the same universal recessive S allele, while the S data Ox24 S1 bb aa bb bb diff Ox25 S2 bb bb bb ab set was derived by coding these unidentified S alleles as Ox26 S4 ab ab bb ab equivalent to the S allele already identified for that Ox27 S5 ab bb bb aa individual, that is, coding all individuals as S homo- zygotes. used to generate the sample individual multilocus genotypes (Table 3). Exact tests of Hardy–Weinberg Results equilibrium genotype frequencies for each of the four allozyme loci indicated random mating with no popula- Allozyme genotyping and population genetic analysis tion substructure in the Oxford S. squalidus sample In all, 13 loci were identified for the nine enzyme systems (Table 4). Furthermore, Fis values for individual and investigated; one locus for each of b-EST, GDH, IDH, combined loci were low, indicating predominant out- 6-PGD, and PGM and two loci for each of AAT, ACO, crossing in the Oxford S. squalidus sample (Table 4). ME, and PGI. Of these loci, seven (AAT-1, AAT-2, ACO- However, an Fis value of 0.28 was measured at the locus 1, ACO-2, b-EST-1, GDH-1, and ME-2) were poly- Aco-1, implying an excess of homozygous genotypes, morphic. Banding patterns at four (AAT-1, ACO-1, b- and permutation tests show the combined locus Fis value EST-1, and ME-2) of these loci were reliably scored and of 0.03 to be significantly greater than zero (Table 4).

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 506 Spatial autocorrelation analysis and at a frequency of 0.06, similar to the expected 0.05 Little evidence for significant SA was found for the Type I error rate for multiple P tests in the absence of SA. allozyme data in any of the combined or individual loci Similarly, Sˇida´k’s test probabilities for each of the analyses carried out (Figure 2; correlograms a and b). correlograms indicate that they could all occur by chance Incidences of significantly positive or negative SA in the absence of SA (Table 5). However, for the first statistics at a 95% confidence level for any particular distance class (0–60 m), a significantly positive Moran’s I locus–distance class combination occurred at random value (I ¼ 0.39, Po0.01) was observed for the Me-2 locus allozyme data indicating at least some SA at this scale. No SA was detectable for either treatment of the S Table 4 Population genetic analysis of Oxford S. squalidus sample locus data as Ssame or Sdiff S genotypes resulted in no allozyme data Moran’s I values significantly greater or less than the expected null value at any of the eight distance classes Statistic Aat-1 Aco-1 b-Est-1 Me-2 All loci (Figure 2; correlograms c, d, and notes).

Frequency of 0.27 NS 0.27 NS. 0.08 NS 0.46 NS — a allele Discussion F 0.07 0.28 0.00 0.00 0.03*. is Our results demonstrate that effective reproduction in NS= nonsignificant w2 test result of Hardy–Weinberg equilibrium British populations of S. squalidus is not impeded by an SSI system operating with as few as six S alleles. In the (H–W exact test). *Significant nonzero Fis value (95% confidence level) based on 1000 repeat permutation tests of combined loci presence of SA, correlograms of Moran’s I typically take allozyme data. the form of significantly positive values in the initial and

Figure 2 Spatial autocorrelation analysis of polymorphic allozyme loci data and S allele data coded as both Ssame and Sdiff data sets for an Oxford S. squalidus sample: (a) individual allozyme loci data, (b) combined allozyme loci data, (c) S locus data coded as Ssame and (d) S locus data coded as Sdiff. Filled or open symbols and long-dash lines are average Moran’s I values for all pairs of plants within each 60 m distance class (see equation (1)). Open symbols and short-dash lines are average 95% conﬁdence limits of Moran’s I values based on random permutations of the data sets for pairs of plants within each 60 m distance class (see Materials and methods). Faint horizontal lines mark out the value for Moran’s I ( ¼ 1/[nÀ1]), expected under the null hypothesis of no spatial autocorrelation. For correlograms a and b representing allozyme loci SA analysis, this value is 0.043 (n ¼ 24), while for correlograms c and d representing S locus SA analysis, it is 0.045 (n ¼ 23). Moran’s I values greater than these indicate positive autocorrelation at a particular distance class, values less than this indicate negative autocorrelation at that distance class (see Materials and methods).

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 507 Table 5 Probabilities that correlograms observed for S. squalidus for substructure in the study population (Table 4). sample population allozyme and S locus data could have occurred Inbreeding in an SI species may be an indirect in the absence of spatial autocorrelation patterns consequence of matings between close relatives, and so should be influenced by local population substructure Correlogram P-value* (Levin and Kerster, 1974; Levin, 1981). In SI S. squalidus, matings between close relatives are possible because of Aat-1 0.14 Aco-1 0.59 the extensive dominance interactions observed between b-Est-1 0.39 S alleles (Brennan et al, 2002). Previous studies of Me-2 0.06 inbreeding in S. squalidus have reported a variety of Fis All loci 0.34 values (0–0.22) using other allozyme loci and popula- Ssame 0.71 tions (Abbott and Forbes, 1993; Abbott et al, 2000). Sdiff 0.62 Significant levels of inbreeding suggest that limited population substructure may be present in natural ˇ *P-values calculated according to Sida´k’s technique (in Materials populations of S. squalidus in Britain. and methods), correlogram significant at a 95% confidence level if The current SA analysis may not have detected 0.025>Po0.975. population substructure if it is present at scales smaller than the 60 m distance classes considered. Distance small distance classes, declining to significantly negative classes below 60 m could not be studied for this sample values at larger distance classes. The Moran’s I correlo- population because of small sample size, necessitating a grams and their 95% confidence intervals for both choice of distance class size containing sufficient num- allozyme and S loci genotype data fail to show any of bers of pairs of datapoints (see Materials and methods). these characteristic patterns associated with SA (Figure 2, This small sample size is an inevitable consequence of correlograms a–d). Not only did these SA analyses find the constraints associated with identifying S locus little evidence for significant population substructure in genotypes through crossing experiments (Brennan et al, the Oxford population of S. squalidus, but they also 2002). showed that mating probably occurs regularly on scales Local relatedness in populations is an important factor of 0–480 m, given the absence of significant negative SA to consider because it may have a large impact on at larger distance classes. reproduction at the population level. For instance, the In particular, the SA analysis of the Ssame and the Sdiff interaction between S-mediated incompatibilities and data sets provided strong evidence for an absence of population substructure has been shown to be respon- spatial structuring of S genotypes in the Oxford popula- sible for correlations between plant proximity and tion of S. squalidus with observed Moran’s I values very reduced fertility in GSI populations of Phlox drummondii close to the value expected under the null hypothesis of (Polemoniaceae; Levin, 1989). In British populations of no SA (Figure 2, correlograms c and d). Neither the Sdiff S. squalidus, incompatibility between individuals due to nor the Ssame ‘genotype’ interpretations of the S genotype local relatedness could be particularly prevalent because data represent plausible patterns of S allele diversity in as few as six S alleles are available for mating SSI populations (Schierup et al, 1997). However, they can interactions (Brennan et al, 2002). Thus, further be interpreted as two alternative ‘genotype’ representa- investigations of small-scale patterns of SA are needed tions of the available S allele data with different chances in natural populations of S. squalidus with suffici- of detecting spatial structuring of S alleles. When all ently dense sampling to allow SA analysis at distances pairs of individuals share at least one S allele in common below 60 m. (Ssame) spatial structure is minimised, whereas when all One potential explanation for the lack of SA in the pairs of individuals share either all or none of their S sample population is that insufficient time has passed, alleles (Sdiff) spatial structure is maximised. The true since the founding of the population, for spatial genetic pattern of spatial structure of the S locus data for the equilibrium to have been achieved. However, this is Oxford population probably lies somewhere between unlikely since simulation studies show that populations these two extremes, neither of which indicated any rapidly develop spatial substructure in the absence of population substructure. homogenising factors within very few generations and Overall, the SA analysis of the allozyme loci data that equilibrium conditions are achieved within about 50 indicates no SA at the scale of the S. squalidus study generations (Sokal and Wartenberg, 1983; Hardy and population corroborating the S locus analysis results. Vekemans, 1999). S. squalidus has been present in Oxford However, the hypothesis that the S locus, under negative for the vast majority of its 300-year history in Britain frequency-dependent selection, should show less SA (Harris, 2002), which is ample time for an equilibrium than more selectively neutral allozyme loci is supported spatial genetic substructure to have developed, even if by the evidence for positive SA at small spatial scales for unfavourable conditions are present. allozyme loci only. Alternatively, the lack of SA in the sample population One locus (Me-2) indicated that there might be positive may be the result of metapopulation dynamics of local SA between S. squalidus individuals at spatial scales of population colonisation and extinction at a rate sufficient less than 60 m. Positive SA at small spatial scales is to disrupt the development of spatial genetic substruc- usually due to increased relatedness between neighbour- ture (Epperson, 1990). S. squalidus occupies highly ing individuals as a consequence of limited pollen and disturbed urban habitats in Britain, so metapopulation seed dispersal (Sokal and Oden, 1978; Sokal and dynamics are an important feature (Brennan personal Wartenberg, 1983). observation). However, colonisation events necessary to Additionally, the small but significantly positive Fis compensate for population extinctions indicate that value (0.03), indicative of inbreeding, is further evidence British S. squalidus populations experience regular long-

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 508 distance gene flow through both pollen and seed. Epperson BK (1990). Spatial autocorrelation of genotypes under Similarly, intrapopulation gene flow is likely to be directional selection. Genetics 124: 757–771. sufficient to maintain adequate mate availability in the Epperson BK (1993). Recent advances in correlation presence of few S alleles. studies of spatial patterns of genetic variation. Evol Biol 27: Reproduction in S. squalidus appears to be little 95–155. affected by low S allele numbers, even though detri- Epperson BK (1995). Fine scale spatial structure: correlations for individual genotypes differ from those for local genotypes. mental effects have been observed in other members of Evolution 49: 1022–1026. the Asteraceae (DeMauro, 1993; Reinartz and Les, 1994; Epperson BK, Huang Z, Li T-Q (1999). Measures of spatial Young et al, 2000) as discussed in the Introduction. The structure in samples of genotypes for multiallelic loci. Genet results of this investigation do not represent a solution to Res Camb 73: 251–261. the problem of low S allele number but rather our data Epperson BK, Li T-Q (1996). Measurement of genetic struc- reinforce the evidence for successful reproduction in S. ture within populations using Moran’s spatial auto- squalidus within the context of a functional SSI system. correlation statistics. Proc Natl Acad Sci USA 93: Many factors could contribute to making 10528–10532. S. squalidus uniquely successful at dealing with low S Epperson BK, Li T-Q (1997). Gene dispersal and spatial genetic allele diversity. Among these factors might be: its recent structure. Evolution 51: 672–681. hybrid origin; non-SSI adaptations to a disturbed Glover BJ, Abbott RJ (1995). Low genetic diversity in the Scottish endemic Primula scotica Hook. New Phytol 129: environment; or even the provision of a new niche after 147–153. introduction to Britain. Tests of these fascinating hypoth- Goudet J (1995). FSTAT (vers. 1.2): a computer program to eses require detailed comparative studies of British S. calculate F-statistics. J Hered 86: 485–486. squalidus’ life history characteristics with these less Hardy OJ, Vekemans X (1999). Isolation by distance in a successful Asteraceae species and also with its ancestral continuous population: reconciliation between spatial auto- parental taxa in Sicily. correlation analysis and population genetics models. Heredity 83: 145–154. Harris SA (2002). Introduction of Oxford Ragwort, Senecio Acknowledgements squalidus L. (Asteraceae), to the United Kingdom. Watsonia 24: 31–43. The work described in this paper was supported Hiscock SJ (2000a). Genetic control of self-incompatibility in primarily by a BBSRC studentship to ACB who grate- Senecio squalidus L. (Asteraceae): a successful colonizing fully acknowledges additional funding from: Depart- species. Heredity 85: 10–19. ment of Plant Sciences, University of Oxford; a Mark Hiscock SJ (2000b). Self-incompatibility in Senecio squalidus L. Quested Award, Fishmongers’ Company, London; a (Asteraceae). Ann Bot 85 (Suppl A): 181–190. Lawrence MJ (2000). Population genetics of the homomorphic Margaret Pollock Award, Somerville College, Oxford; SI polymorphism in flowering plants. Ann Bot 85 (Suppl A): and The Lawlor Foundation, Belfast. Thanks also to: Phil 221–226. Smith for care and maintenance of greenhouse plants; Dr Levin DA (1981). Dispersal versus gene flow in plants. Ann M John Brookfield; Dr Richard Abbott; and two anonymous Bot Gard 68: 233–253. reviewers who provided constructive criticisms to earlier Levin DA (1989). Proximity-dependent cross-compatibility in drafts of this paper. Phlox. Evolution 43: 1114–1116. Levin DA (1996). The evolutionary significance of pseudo-self- fertility. Am Nat 148: 321–332. References Levin DA, Kerster HW (1974). Gene flow in seed plants. Evol Biol 7: 139–220. Abbott RJ, James JK, Irwin JA, Comes HP (2000). Hybrid origin Manly BFJ (1997). Randomization Bootstrap and Monte Carlo of the Oxford Ragwort, Senecio squalidus L. Watsonia 23: Methods in Biology. Chapman & Hall, London. 123–138. Oden NL (1984). Assessing the significance of a spatial Abbott RS, Forbes DG (1993). Outcrossing rate and self- correlogram. Geogr. Anal. 16: 1–16. incompatibility in the colonising species Senecio squalidus. Raymond M, Rousset F (1995). GENEPOP (version 1.2): Heredity 71: 155–159. population genetics software for exact tests and ecumeni- Ashton P (1990). Multiple origins of Senecio cambrensis Rosser cism. J Hered 86: 248–249. and related evolutionary studies in British Senecio. PhD Reinartz JA, Les DH (1994). Bottleneck induced dissolu- Thesis, University of St Andrews. tion of self-incompatibility and breeding system conse- Bacilieri R, Labbe T, Kremer A (1994). Intraspecific genetic quences in Aster furcatus (Asteraceae). Am J Bot 81: structure in a mixed population of Quercus petraea (Matl.) 446–455. Leibl and Q. robur L. Heredity 73: 130–141. Richman AD, Uyenoyema MK, Kohn JR (1996). Allelic diversity Brennan AC, Harris SA, Tabah DA, Hiscock SJ (2002). The and gene genealogy at the self-incompatibility locus in the population genetics of sporophytic self-incompatibility in Solanaceae. Science 273: 1212–1216. Senecio squalidus L. (Asteraceae) I: S allele diversity in a Schierup MH, Vekemans X, Christiansen FB (1997). Evolution- natural population. Heredity 89: 430–438. ary dynamics of self-incompatibility alleles in plants. Genetics Byers DL, Meagher TR (1992). Mate availability in small 147: 835–846. populations of plant species with homomorphic self-incom- Sokal RR, Oden NL (1978). Spatial autocorrelation in biology 1. patibility. Heredity 68: 353–359. Methodology. Biol J Linn Soc 10: 199–228. Degen B, Petit R, Kremer A (2001). SGS – Spatial Genetic Sokal RR, Wartenberg DE (1983). A test of spatial autocorrela- Software: a computer program for analysis of spatial genetic tion analysis using an isolation-by-distance model. Genetics and phenotypic structures of individuals and populations. J 105: 219–237. Hered 92: 447–448. Streiff R, Labbe T, Bacillieri R, Steinkeller H, Glossl J, Kremer A DeMauro MM (1993). Relationship of breeding system to rarity (1998). Within-population genetic structure in Quercus robur in the lakeside daisy (Hymenoxys acaulis var. glabra). Conserv L. and Quercus petraea (Matt.) Liebl. assessed with isozymes Biol 7: 542–550. and microsatellites. Mol Ecol 7: 317–328.

Heredity Population genetics of SSI in Scenecio squalidus AC Brennan et al 509 Waser NM, Mitchell RJ (1990). Nectar standing crops in Wright S (1939). The distribution of self-sterility alleles in Delphinium nelsonii ﬂowers: spatial autocorrelation among populations. Genetics 24: 538–552. plants? Ecology 71: 116–123. Young AG, Miller BG, Gregory E, Langton A (2000). Sporo- Wendel JF, Weeden NF (1990). Visualisation and interpretation phytic self-incompatibility in diploid and tetraploid races of plant isozymes. In: Soltis DE, Soltis PS (eds) Isozymes in of Rutidosis leptorrhynchoides (Asteraceae). Aust J Bot 48: Plant Biology, Chapman & Hall: London. p 5–45. 667–672.

Heredity